Special Session 57: Mathematical models for traffic monitoring and control

Using a Bi-level optimization algorithm to calibrate traffic simulations for the CIRCLES experiment

Sean T McQuade
Rutgers-Camden
USA
Co-Author(s):    
Abstract:
The CIRCLES consortium performed the largest live traffic control experiment to smooth traffic waves during rush hour on Interstate 24 West heading into Nashville, TN. It was performed during the week of November 14, 2022. To plan for the 100 vehicle test, a Bi-level optimization for an agent-based dynamic traffic assignment model (DTA) was used to calibrate a simulator. The supply parameters, such as traffic network topology, capacity, and speed limits, define the environment. The demand inputs represent the travelers and their behavior, such as origin-destination matrices, routing, and lane changing. The environment was calibrated to the background traffic where we ran the live experiment, using flow-speed data. Then, the control vehicles were added to the simulator to estimate characteristics of their drive in the calibrated background traffic, such as how they would affect traffic on arterial roads, and how they may cluster at off ramps.